A Machine Learning Approach to Quantifying Observation Error of Airborne Radio Occultation in Atmospheric Rivers

Aug 5, 2026

Upload by Eden

This thesis presents a machine-learning framework to quantify observational error in airborne radio occultation (ARO) measurements deployed over atmospheric rivers (ARs). It examines the synoptic environment relevant to ARs, describes the ARO data collected during AR reconnaissance, and reviews ARO retrieval techniques along with major sources of observation error and quality-control procedures. The study compares ARO results with ERA5 reanalysis, then uses clustering and Gaussian mixture model methods to characterize vertical AR structure, assign cluster labels, and evaluate ARO–ERA5 differences across clusters and geographic regimes.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages85

Reading Time3 h 34 min

AccessFree to download

Chapter 1 Introduction

Chapter 2 Synoptic environment in and near ARs

Chapter 3 ARO data collected during AR Recon

Chapter 4 Methods

4.1 ARO Retrieval Techniques

4.2 Sources of ARO Observation Error

4.3 Quality Control for the Purposes of this Study

4.4 Simulation of Refractivity from Model Products

4.5 Simulation of bending angle from model products

Chapter 5 Comparing ARO and ERA5 reanalysis

5.1 Aggregate statistics of ARO-ERA5 differences

5.2 ARO-ERA5 differences as a function of latitude

5.3 ARO-ERA5 differences near the tracking limit

Chapter 6 Utilizing machine learning to characterize the vertical structure of the AR environment

Chapter 7 Results of selected GMM method

Chapter 8 Discussion

UC San Diego UC San Diego Electronic Theses and Dissertations Title A Machine Learning Approach to Quantifying Observation Error of Airborne Radio Occultation in Atmospheric Rivers Permalink [https://escholarship.org/uc/item/2pr2z3g8](https://escholarship.org/uc/item/2pr2z3g8) Author Barton, Noah Jeffery Publication Date 2025 Peer reviewed|Thesis/dissertation [eScholarship.org](eScholarship.org) Powered by the California Digital Library University of California UNIVERSITY OF CALIFORNIA SAN DIEGO A Machine Learning Approach to Quantifying Observation Error of Airborne Radio Occultation in Atmospheric Rivers A Thesis submitted in partial satisfaction of the requirements for the degree Master of Science in Oceanography by Noah Jeffery Barton Committee in charge: Professor Jennifer S. Haase, Chair Professor Joel Norris Professor Duncan Watson-Parris Professor Shang-Ping Xie Copyright Noah Jeffery Barton, 2025 All rights reserved. The Thesis of Noah Jeffery Barton is approved, and it is acceptable in quality and form for publication on microfilm and electronically. University of California San Diego 2025 TABLE OF CONTENTS Thesis Approval Page ......................................................... iii Table of Contents ............................................................ iv [List of Figures ............................................................... vi](List of Figures ............................................................... vi) [List of Tables ................................................................ vii](List of Tables ................................................................ vii) [Acknowledgements ........................................................... viii](Acknowledgements ........................................................... viii) [Abstract of the Thesis ........................................................ xi](Abstract of the Thesis ........................................................ xi) [Chapter 1 Introduction ..................................................... 1](Chapter 1 Introduction ..................................................... 1) [Chapter 2 Synoptic environment in and near ARs .............................. 5](Chapter 2 Synoptic environment in and near ARs .............................. 5) [Chapter 3 ARO data collected during AR Recon ............................... 9](Chapter 3 ARO data collected during AR Recon ............................... 9) [Chapter 4 Methods ........................................................ 15](Chapter 4 Methods ........................................................ 15) [4.1 ARO Retrieval Techniques ............................................. 15](4.1 ARO Retrieval Techniques ............................................. 15) [4.2 Sources of ARO Observation Error ...................................... 16](4.2 Sources of ARO Observation Error ...................................... 16) [4.3 Quality Control for the Purposes of this Study ............................. 17](4.3 Quality Control for the Purposes of this Study ............................. 17) [4.4 Simulation of Refractivity from Model Products ........................... 18](4.4 Simulation of Refractivity from Model Products ........................... 18) [4.5 Simulation of bending angle from model products ......................... 20](4.5 Simulation of bending angle from model products ......................... 20) [Chapter 5 Comparing ARO and ERA5 reanalysis ............................... 21](Chapter 5 Comparing ARO and ERA5 reanalysis ............................... 21) [5.1 Aggregate statistics of ARO-ERA5 differences ........................... 21](5.1 Aggregate statistics of ARO-ERA5 differences ........................... 21) [5.2 ARO-ERA5 differences as a function of latitude .......................... 24](5.2 ARO-ERA5 differences as a function of latitude

FAQs

What is the main goal of the thesis?

To quantify observation error in airborne radio occultation measurements over atmospheric rivers using machine-learning methods and comparative analysis with reanalysis data.

How does the thesis assess observation error?

It identifies sources of ARO observation error, applies quality control, simulates refractivity and bending angles from model products, and compares ARO with ERA5 reanalysis.

What machine-learning approach is used to characterize the atmospheric river environment?

The study applies clustering and Gaussian mixture model techniques to learn vertical structure patterns, assign ARO profiles to clusters, and relate cluster membership to ARO–ERA5 differences across locations and regimes.

Aug 5, 2026

Upload by Eden